nonprofit groups like foundations pay the initial money for a new program and also oversee it, with government approval. The government will reimburse them several years later, possibly with a bonus -- but only if agreed-upon benchmarks show that the program is working.
The advantage of this approach is that programs that fail will run out of money. For example, imagine that a charter school was funded by social investors who will be reimbursed by the government only if the school meets benchmarks. If the school fails to meet those benchmarks, the investors will not get paid back and they probably will close the school. That is a good thing.
It is interesting to me that this approach is limited to nonprofits. Why should it make a difference to the government whether the organizations that successfully meet objectives are nonprofit or for-profit?
Of course, I think the overall notion is a crock. It is unlikely that bureaucrats in Washington are going to have sufficient information to design and implement a compensation scheme that works effectively. If that were possible, then we would observe successful examples of governments using market socialism. Instead, I think that Hayek is right, and the importance of local knowledge makes this form of central planning ineffective. At best, it would be less ineffective than other forms of central planning.